Allow selection based on FDR q-value

[legaultmarc]

Instead of using a p-value threshold, it would be nice to allow selection based on the FDR.

[legaultmarc]

Using this approach would require accounting for LD or clumping variants first.

[legaultmarc]

Actually an approach similar to what is used right now but that accumulates the p-values of the selected variants until FDR is reach would be pretty easy to do.